Sei sulla pagina 1di 45

HEAT TRANSFER

FINITE ELEMENT FORMULATION

VIJAYAVITHAL BONGALE
DEPARTMENT OF MECHANICAL ENGINEERING
MALNAD COLLEGE OF ENGINEERING
HASSAN - 573 202.
Mobile : 9448821954

17/08/2014 N I T CALICUT
17/08/2014 N I T CALICUT
What is Finite Element Method?

 FEM is a numerical analysis technique


for obtaining approximate solutions to
a wide variety of engineering
problems.

17/08/2014 N I T CALICUT
Applications of FEM:
1. Equilibrium problems or time independent problems.
e.g. i) To find displacement distribution and stress
distribution for a mechanical or thermal loading in
solid mechanics. ii) To find pressure, velocity,
temperature, and density distributions of equilibrium
problems in fluid mechanics.
2. Eigenvalue problems of solid and fluid mechanics.
e.g. i) Determination of natural frequencies and
modes of vibration of solids and fluids. ii) Stability
of structures and the stability of laminar flows.
3.Time-dependent or propagation problems of continuum
mechanics.
e.g. This category is composed of the problems that
results when the time dimension is added to the
problems of the first two categories.

17/08/2014 N I T CALICUT
Similarities that exists between various
types of engineering problems:
1. Solid Bar under Axial Load

  u 
 AE   0,
x  x 
Where , E is the Young' s modulus,
u is axial displacement,
and A is cross  sectional area
17/08/2014 N I T CALICUT
2. One – dimensional Heat Transfer
  T 
 KA   0, Laplace equation
x  x 
Where , K is the thermal conductivity,
T is temperatur e, and A is cross  sectional area
3. One dimensional fluid flow
  Φ 
 ρA   0, Where ,
x  x 
ρ is the density , φ  is potential function ,
Φ
and A is cross  sectional area and u 
17/08/2014 N I T CALICUT
x
How the Finite Element Method Works
 Discretize the continuum: Divide the
continuum or solution region into elements.
 Select interpolation functions: Assign nodes
to each element and then choose the
interpolation function to represent the variation
of the field variable over the element.
 Find the Element Properties: Determine the
matrix equations expressing the properties of
the individual elements. For this one of the three
approaches can be used. i) The direct approach
ii) The variational approach or iii) the weighted
residuals approach.
17/08/2014 N I T CALICUT
 Assemble the Element Properties to Obtain the
System equations: Combine the matrix equations
expressing the behavior of the elements and form the
matrix equations expressing the behavior of the entire
system.
 Impose the Boundary Conditions: Before the
system equations are ready for solution they must be
modified to account for the boundary conditions of the
problem.
 Solve the System Equations: Solve the System
Equations to obtain the unknown nodal values like
displacement, temperature etc.
 Make Additional Computations If Desired: From
displacements calculate element strains and stresses,
from temperatures calculate heat fluxes if required.
17/08/2014 N I T CALICUT
17/08/2014 N I T CALICUT
General Steps to be followed while solving a
problem on heat transfer by using FEM:

1. Discretize and select the element type


2. Choose a temperature function
3. Define the temperature gradient / temperature
and heat flux/ temperature gradient relationships
4. Derive the element conduction matrix and
equations by using either variational approach or
by using Galerkin’s approach
5. Assemble the element equations to obtain the
global equations and introduce boundary
conditions
6. Solve for the nodal temperatures
7. Solve for the element temperature gradients and
heat fluxes
17/08/2014 N I T CALICUT
ONE DIMENSIONAL FINITE ELEMENT
FORMULATION USING VARIATIONAL
APPROACH
Step 1.Select element type. Step 2.Choose a temperature
function.
x
t1 L t2 T = N1t1+ N2 t2

1 2 t2

One –D element t1

1 2

Temperature Variation along the


length of element

17/08/2014 N I T CALICUT
We choose, T (x) = N1 t1 + N2 t2 --------------------------- (1)

N1 & N2 are shape functions given by,

x x
N1  1  , N2     (2)
L L
In matrix form

 x x t1 
N  1   and t         (3)
 L L t 2 
and

T   N 
t      (4)

17/08/2014 N I T CALICUT
Step 3. Define the temperature gradient / temperature and heat flux/

temperature gradient relationships

Temperature gradient matrix is given by,

 dT 
g     Bt          (5)
 dx 
Where matrix B is,

 dN1 dN 2   1 1 
B                (6)
 dx dx   L L 
The heat flux /temperature gradient relationship is,

q x  Dg         (7)
17/08/2014 N I T CALICUT
The material property matrix is,

D  K xx         (8)
Step 4. Derive the element conduction matrix and equations

Consider the following equations


2 T=TB
d T
K xx 2
 Q  0        (I) qx=0
dx
q x = + q*x S1
L

T=TB

Insulated

L S2
S1
With T =TB on surface S1,
dT
17/08/2014 q  K xx
*
N I T CALICUT
x  Constant on S 2
dx
Where, Q is heat generated / unit volume
*
q x is the heat flow / unit area
is positive when heat is flowing into body
is negative when heat is flowing out of the body
is Zero on an insulated boundary
Consider
  T  T hP
 K xx   Q  ρC  T  T       (II)
x  x  t A
With the first boundary condition
h
of above equation and /or second
boundary condition and /or loss of
heat by convection from the ends
T T of 1-D body, we have
2

 h T  T  on surface S3
dT
Insulated
17/08/2014 I TK
N xx
CALICUT
dx
Minimize the following functional :
(Analogous to the potential energy functional Π)

Π h  U  Ω Q  Ω q  Ω h            III
Where,

1   dT  
2

U   K xx    dV
2 v   dx  

Ω Q    Q T dV Ωq    q * T dS
V S2

Ω h   hT  T  dS          (9)
1 2

17/08/2014
2 S3 N I T CALICUT
Important:
q* and h on the same surface cannot be specified simultaneously
because they cannot occur on the same surface

We now have the functional given by


Πh  U  ΩQ  Ωq  Ωh
1   dT  
2
1
  K xx    dV   Q T dV   q T dS   h T  T  dS    (10)
* 2

2 v   dx   V S 2
2S 3

Consider the first term,

1    
 
2

K xx    dV   g Dg dV          (11)


dT 1

T

2 v   dx   2V

17/08/2014 N I T CALICUT
Second term gives,

 Q T dV   t N  Q dV - - - - - - - (12)
T T

V V

Third term gives,

 T dS 
q *
 t T
N  q dS        (13)
T *

S2 S2

Similarily, fourth term gives,

1
2S h T  T

2
dS 
1
2S h t 
T
N  T
 T  2
dS        (14)
3 3

Substituting
17/08/2014
equations (11),(12),(13) and (14) in equation (10) we obtain
N I T CALICUT
2V
T
 T

Π h   g Dg dV   t N  Q dV
1 T

T T *

2S
T T

  t N  q dS   h t N   T dS
1 2

S 2 3

 t  B DBdVt  t  N  Q dV  t  N  q * dS


1 T T T T T T

2 V V S 2

1
2S
T T
 T
 T
 
  h t N  N t  t N   N t T  T2 dS    (15)
3

Equation (15) has to be minimized with respect to t 


and equated to zero

17/08/2014 N I T CALICUT
Π h
  B DBdV t   N  Q dV   N  q * dS
T T T

t V V S 2

  h N  N  dS t   N  hT dS  0        (16)


T T

S3 S3

On simplifying,

 
  B DBdV   h N  N  dS  t  f Q  f q  f h     (17)
T T

V S 3 

The above equation is of the form,

f   K  t        (18)
17/08/2014 N I T CALICUT
Where,
Element Conduction matrix is
K    B DBdV   h N N dS K   K         (19)
T T
C h
V S3

The first term is conduction part of K and


second term represent convection part of K

And the force matrices have been defined by,


f    N Q dV ,
Q
T

f   N q *dS,
q
T

S2

f    N 
17/08/2014h
T
hT dS N I T CALICUT
S3
Consider the conduction part,

K    B DBdV - - - - - - - (20)


C
T

V
Substituting for B, D and dV in the above equation,

 1
L
 
 L  1 1
0  1  K xx   L L  A dx
 

 L 
A K xx L  1  1  1  1
dx , i.e. K C  
A K xx
  
L 0  1 1 
2   1 1       (21)
L  

17/08/2014 N I T CALICUT
The convection part is,

 x
L
1 
hPL 2 1 
K h    h NT N dS  hP   L 1  x x
 dx         (22)
S3 0 x  L L 6 1 2
 L 
Where dS P dx

Therefore, element conduction matrix is,

 1  1 hPL 2 1 
K  A K xx
 1 1   6 1 2      (23)
L    

17/08/2014 N I T CALICUT
The force matrix terms will be,

 x
L
1 
 L QAL 1
f Q    N Q dV  QA   dx 
T
,
V 0  x  2 1
 L 
 x
L
1 
 L q * PL 1
f q   N q * dS  q * P    dx 
T

S 2 0  x  2 1
 L 
hT PL 1
f h    N hT dS 
T

S 3
2 1
17/08/2014 N I T CALICUT
By adding we get,

QAL 1 q * PL 1 hT PL 1


f   f Q   f q   f h       
2 1 2 1 2 1
QAL  q * PL  hT PL 1
            (24)
2 1
Convection force from the end of the element

T00

1 2

17/08/2014 N I T CALICUT
We have an additional convection term contribution to
the stiffness matrix and is,

K  h end
  h N N dS        (25)
T

Send

N1 = 0 and N2 = 1 at right end and

0 0 0
K 
h end
  h   0 1dS  hA          (26)
Send 1  0 1
The convection force from the free end

 N1 ( x  L)  0
f   hT A   hT A       (26)
 N 2 ( x  L) 
h end
1 
17/08/2014 N I T CALICUT
Step 5. Assemble the element equations to obtain the
global equations and introduce boundary conditions.

The global structure conduction matrix is

K    K 
n
e

e 1

The global force matrix is

F   f 
n
e

e 1
and global equations are

17/08/2014
F   K t
N I T CALICUT
Step 6. Solve for the nodal temperatures

Step 7. Solve for the element temperature gradients and


heat fluxes.

17/08/2014 N I T CALICUT
ONE DIMENSIONAL FINITE ELEMENT FORMULATION
GALERKIN’S APPROACH

Equation representing one dimensional formulation of


conduction with convection is given by,
d 2T hP
kx 2  Q  T  T       (1)
dx A
Element – 1-D linear 2 noded element with temperature
function
T (x) = N1 t1 + N2 t2 --------------------------- (2)
Where,
x x
N1  1  , N 2 
17/08/2014 L L
N I T CALICUT
The residual equations for the equation (I) are
x2
 d 2T hP 
x k x dx 2  Q  A T  T  N i A dx  0      (3)
1  
x2
d 2T x 2 x 2 x 2

A  k x 2 N i dx  hP  T N i dx  A  Q N i dx  hP T  N i dx  0  (4)
x1
dx x 1 x 1 x 1

Where i = 1,2

Integrating the first term of equation (4) by parts and


rearranging,
x2 x2 x2 x2 x2
dNi dT dT
kxA  dx  hP  T N i dx  A  Q N i dx  hP T  N i dx  K x AN i  (5)
x1 dx dx x 1 x 1 x1
dx x2

17/08/2014 N I T CALICUT
Integration by parts:

 udv  uv   vdu
dNi
Where u  N i , du  dx ,
dx
d  dT  dT
dv   A K xx  dx and v  A K xx
dx  dx  dx

17/08/2014 N I T CALICUT
Substituting for T , we get

dNi  dN1
x2
dN2  x

t 2  dx  hP  N i N1t1  N 2 t 2  dx
2

kxA   t1 
x dx  dx
1
dx  x 1

x2 x2 x2
dT
 A  Q N i dx  hP T  N i dx  K x AN i  (6)
x1 x1 dx x2

Or
T
 dN   dN 
x2 x

t dx  hP  N Nt dx


2

kxA  
T
  
x  dx   dx 
1 x 1

 A  Q N 
x2 x2 x2

dx  hP T  N dx  K x AN
dT
 (7)
T T T

x1 x 1
dx x2

17/08/2014 N I T CALICUT
The above equations are of the form,

k t  f  f  f      (8)
e e
Q
e
q
e
h

And
T
   
k   k x A     dx  hP  N N dx
x2 x
e
dN dN T
2

x  dx   dx 
1 x 1

 k   k     (9)
e
c
e
h

Let x1= 0 and x2 = L, then,


17/08/2014 N I T CALICUT
If one substitute for N1 & N2 in equation (9) and
solve, then,

 1  1 hPL 2 1 
k eA kx
 1 1    
L   6 1 2

 k   k      (10)
e
c
e
h

The forcing function vectors on the right hand


side of the equation (7) are given by
17/08/2014 N I T CALICUT
L 
0 QN1 dx  QAL 1
f Q   A  L
e
 
 QN dx 2 1
0 2

 dT 
 dx 
f 
e
q

 kxA 
0  q x 0 
  A
 q1 
  A 
 dT  - q x L  - q 2 
 dx L 

L 
0 N1dx  hT PL 1
f h   hT P  L 

         (11)
 N dx 2 1
17/08/2014 0 2  N I T CALICUT
Two Dimensional Finite Element
Formulation:

 1-d elements are lines


 2-d elements are either triangles,
quadrilaterals, or a mixture as
shown
 Label the nodes so that the
difference between two nodes on
any element is minimized.

17/08/2014 N I T CALICUT
Three noded triangular element:

Assume (Choose) a Temperature Function:


TNt Nt N t
i i j j m m

t 
 
T   N N N t
i

i 
j m j

t 
  m

t  nodal temperature

17/08/2014 N I T CALICUT
Define Temperature Gradient Relationships
 T   N N N  t
j

 x   x 
i m

x x  Analogous to strain
i

g       t  matrix: {g}=[B]{t}



   N
T i
N j
N m 
j


 y   y y y   t m 

Where, N 
1
α i i
 β x  γ y i i
2A
N 
1
α j j
 β x  γ yj j
2A
N 
1
α m m
 β x  γ y m m
2A
and α  x y  y i j m j
x , β y  y , γ x x
m i j m i m j

α x y y x , β y  y , γ x x
j m i m i j m i j i m

17/08/2014 α  x y  y x , β N I y
m i
 y , γ x x
T CALICUT
j i j m i j m j i
[B] is derivative of [N]:

 1    
B    N    
i j m

x 2 A     i j m

Heat flux/ temperature gradient relationship is:

q K 0 
   
x xx

g  Dg
q y 0 K  yy

K 0 
Where D  
xx


 0 K  yy

17/08/2014 N I T CALICUT
Derive the element conduction matrix and equations:
K   B  DB dV  h  N   N  dS  K   K 
T T
 
V S3 C h

Where,
  
1    K 0     
i i

K    B DBdV       
T

  dV
xx i j m

0 K     
C j j


2
V
2A V

   yy i j m

  m m

If thickness is assumed constant and all terms of


integrand constant, then the conduction portion of
the total stiffness matrix is,
K   B DBdV  tAB DB
C

V
T T

17/08/2014 N I T CALICUT
Now the convection portion of the total stiffness
matrix is,
K   h N  N  dS
h

S3
T

N N N N N N 
 
i i i j i m

 h  N N N N N N
S3 j i j j j m  dS
N N N N N N 
 m i m j m m 
Consider the side between nodes i
and j of the element subjected to
convection, then, Nm = 0 along side i-j

17/08/2014 N I T CALICUT
We Obtain 2 1 0 
hL t  
K  i j
120
6  
h

0 0 0 
Where Li-j is the length of side i-j

Force Matrices:

 f   N  Q dV  Q N  dV
Q

V
T

V
T

1
QV  
 1
3  
17/08/2014
1 N I T CALICUT
 f 
q

S2
q *  N  dS
T

N
 1 
 q* L t  
i

  q *  N
 dS  j 1  on side i  j
i- j

S2

N
 2  
 m 0 
0  0 
q* L t   q* L t  
 1  on side j  m 
j -m
1  on side m  i
m -i

2   2  
1  1 
The integral
f  h T  N  dS
T

h S3 

Can be found in a same manner by simply replacing


17/08/2014 N I T CALICUT q * with h T

 f  h T  N  dS
T

h S3 

N  1 
  hT L t  
i

 hT  N
  dS 
 j  1  on side i  j
 i- j

S2

N  2 0 
  m  
0 
hT L t  
 
 1  on side j  m
j -m

2 1 
 
0 
hT L t  
 
 1  on side m  i
m -i

2 1 
 

17/08/2014 N I T CALICUT
Thank You

17/08/2014 N I T CALICUT

Potrebbero piacerti anche